基于LDA主题模型的协同过滤推荐算法

Collaborative filtering recommendation algorithm based on LDA topic model

传统的协同过滤推荐算法直接根据用户对物品的评分进行推荐,忽略了评论文本中隐含的重要信息,当用户对物品的评论较少时,由于数据的稀疏性会造成推荐效果的不准确和单一。本文提出了一种基于LDA主题模型的协同过滤推荐算法LDA-CF(Latent Dirichlet Allocation model-LDA-Collaborative Filtering),在传统的协同过滤算法基础上,通过LDA模型对评论文本中的主题进行分类,从各个主题层面挖掘用户的情感偏好,计算用户之间的相似度,进而向目标用户推荐商品。对京东平台牙膏的评论数据集的实验结果表明,该算法不仅可以缓解由于评分数据较少造成的稀疏性问题,推荐的精确度也有所提高。

Traditional collaborative filtering recommendation algorithms tend to recommend items directly according to users′scores,ignoring the important information implied in the comment text.Moreover,when users have few comments on items,the sparsity of the data will lead to the inaccuracy and singleness of the recommendation effect.Therefore,this paper proposes a collaborative filtering recommendation algorithm based on LDA topic model.Based on the traditional collaborative filtering algorithm,the algorithm classifies the topics in the review text through the LDA model,mines the emotional preferences of users from each topic level,calculates the similarity between users,and then recommends products to target users.The experimental results based on the review data set of toothpaste on JD platform show that the algorithm can not only alleviate the sparsity problem caused by few score data,but also improve the recommendation accuracy compared with the traditional collaborative filtering algorithm.